Manual setup required. The maintainer's config contains paths only you know - edit the placeholders below before adding it to Claude Code.
1
Prepare the server locally
Run this once before adding it to Claude Code.
git clone https://github.com/zheng/crag.git
cd crag
go install ./cmd/crag2
Register it in Claude Code
claude mcp add code-rag-golang -- crag mcp -d /absolute/path/.crag.dbReplace any placeholder paths in the command with the real path on your machine.
3
Make your agent remember this setup
code-rag-golang's config, env vars, and the gotchas you hit — recalled in every future Claude Code, Cursor, and Codex session.
npx conare@latestFree · one command · indexes the sessions already on disk. Set up in the browser instead →
What it does
- Builds precise call graphs using SSA and VTA static analysis
- Persists analysis data in a queryable SQLite database
- Provides impact analysis to identify affected callers before code changes
- Supports incremental updates via file watching or git hooks
- Exports call graph data as Markdown for RAG context
Tools 6
impactPerform impact analysis to find callers and callees of a function.upstreamFind recursive upstream callers of a function.downstreamFind what a specific function calls.searchSearch for functions by name.riskShow high-risk functions in the codebase.implementsList interface implementations.Try it
→Where is the HandleRequest function called in this project?
→If I modify the BuildSSA function, what other parts of the code are affected?
→Find all functions in the codebase that contain the word 'Auth'.
→List all high-risk functions identified by the static analysis.
→Show me all implementations of the current interface.
Original README from nullptr-z/code-rag-golang
crag - Call Graph for AI-Assisted Go Development
AI modifies a function but misses 5 callers that also need updating. crag fixes this — it builds precise call graphs via static analysis, so AI knows exactly what's affected before making changes.
How It Works
crag analyze . # Build call graph using SSA + VTA static analysis
# → Stored in SQLite, fast to query
AI: "Modify ProcessRequest"
→ crag automatically finds all 5 callers
→ AI updates them together, nothing missed
Install
git clone https://github.com/zheng/crag.git
cd crag
go install ./cmd/crag
Quick Start
# 1. Analyze your Go project
crag analyze . -o .crag.db
# 2. Configure MCP for your AI editor (Cursor / Claude Code)
# Add to .cursor/mcp.json or claude_desktop_config.json:
{
"mcpServers": {
"crag": {
"command": "crag",
"args": ["mcp", "-d", "/absolute/path/.crag.db"]
}
}
}
# 3. Keep it updated (pick one)
crag watch . -d .crag.db # Auto-update on file changes
# or: add `crag analyze . -i` to .git/hooks/post-commit
That's it. Your AI editor can now query call graphs directly.
What AI Can Do With crag
After MCP setup, just ask naturally:
- "Where is HandleRequest called?" → upstream callers
- "If I change BuildSSA, what's affected?" → impact analysis
- "Find all functions containing Auth" → search
CLI Usage
crag impact "HandleRequest" -d .crag.db # Impact analysis (callers + callees)
crag upstream "db.Query" -d .crag.db # Who calls this? (recursive)
crag downstream "Process" -d .crag.db # What does this call?
crag search "Handler" -d .crag.db # Search functions by name
crag risk -d .crag.db # Show high-risk functions
crag implements -d .crag.db # Interface implementations
crag view -d .crag.db # Web UI visualization
crag export -d .crag.db -o crag.md # Export as Markdown (RAG context)
Why crag?
| Text search (grep) | IDE (gopls) | crag | |
|---|---|---|---|
| Interface calls | miss | partial | VTA precise resolution |
| Persisted & queryable | no | no | SQLite |
| AI integration | manual copy | no | MCP native |
| Incremental update | n/a | n/a | Git-aware |
| Zero CGO | n/a | n/a | Pure Go SQLite |
Tech
- Analysis: Go SSA + VTA (Variable Type Analysis) via
golang.org/x/tools - Storage:
modernc.org/sqlite(pure Go, single binary) - CLI:
cobra· Web UI: embeddedvis.js
License
MIT